This dataverse contains the Struct-BB Dataset, an annotated image dataset of
exposed reinforcement in bridges and buildings, designed for object detection
and segmentation tasks using deep learning models.

The data are organized into image files and annotation files compatible with
YOLO-based frameworks.

Also included are pixel-level segmentation masks corresponding to the segmented
images.

The data set names are:

base_images_masks/
base_images_original/
base_images_segmented/
base_labels/
aug_images_masks/
aug_images_original/
aug_images_segmented/
aug_labels/
data_documentation.txt

Directory descriptions:

base_images_masks
Contains binary pixel-level segmentation masks used to generate the images in
the images_segmented folder (.png)

base_images_original
Contains the original RGB images collected from field inspections of bridges
and buildings under real-world conditions (.jpg)

base_images_segmented
Contains images generated based on the corresponding segmentation masks (.jpg)

base_labels
Contains annotation files in YOLO format (.txt), with normalized coordinates
for object detection and segmentation.

aug_images_original
Contains augmented RGB images generated from the original images through data augmentation techniques to increase data diversity for model training (.jpg)

aug_images_masks
Contains augmented binary segmentation masks (.png)

aug_images_segmented
Contains augmented segmented images (.jpg)

aug_labels
Contains annotation files in YOLO format for augmented images (.txt)

Data documentation file:

data_documentation.txt
Contains image-level metadata linking each RGB image to its corresponding mask
and YOLO annotation file, including the number of annotated instances, defect
class, environmental conditions, and preprocessing information.

The dataset includes one annotated class:

0 - Exposed reinforcement